LoWait: Learning Optimal Waiting Time Threshold to Minimize Age of Information Over Wireless Fading Channels

This paper focuses on using the age of information (AoI) as a key metric for measuring the freshness of status information in a wireless fading channel. Specifically, we examine a scenario where a sensor can sample, generate, and transmit update packets at will, with the stochasticity of transmission rate intricately linked to the channel state. Moreover, our goal is to formulate a waiting policy that minimizes the time average AoI by determining the waiting time following the completion of a transmission. We first introduce a straightforward yet efficient Learning Optimal Waiting Time Threshold (LoWait) algorithm for two-state Markov channels. LoWait operates without the need for prior knowledge of the probability distributions for transmission rates and the parameters of channel alternation. In the classical ON-OFF channels, we introduce the Robbins-Monroe iteration and prove that the optimality gap of LoWait diminishes at a rate of
${\mathrm {O}}(1/\sqrt{K})$
, where K is the number of updates. Numerical results affirm that LoWait can effectively reduce the average AoI compared to other benchmarks.